Catálogo de publicaciones - libros

In this paper, we introduce an information hiding technique into remote sensing area. We develop its connotation that the secret information is still hidden in the original remote sensing image. We propose a practical information hiding technique and a novel wavelet information hiding algorithm which is able to adapt to features of a remote sensing image. The technique is based on the embedding strategy of Discrete Wavelet Transform and HVS (Human Visual System) character. The algorithm is a blind one and has no influence on applied value of a remote sensing image.

It is a challenging issue to analyze video content for video mining due to the difficulty in video representation. A hierarchical model of video representation is proposed with a schema for content-based analysis of news video in this paper. The research problem targeted in this paper is to mine a massive video database to retrieve specific clip based on content defined by users. This is frequently encountered in entertainment and video editing. A novel solution to this problem is developed in this paper, in which the consecutive news video is segmented into shots, scenes and news items using multimodal features based on the hierarchical model. To summarize the content of video, a video abstract is developed. The experimental evaluation demonstrates the effectiveness of the approaches discussed in this paper.

A coarse-to-fine automatic point-based image registration method is proposed in this paper. At the first stage, clustering is used to determine the scale parameter and the rotational parameter candidates between images. Convex hull vertices correlation is applied subsequently to determine the correct rotational parameter. With the coordinates of matched point pairs and the above parameters, the translational parameter and the coarse registration result can be determined. At the second stage, control point pairs, which determine parameters of mapping polynomial, are formed by iterative convex hull vertices matching. Thus the registration result is refined. Experiments indicate that this approach can automatically align images in different resolutions.

An important step in automatic fingerprint recognition systems is the segmentation of fingerprint images. In this paper, we present an adaptive algorithm based on Gaussian-Hermite moments for non-uniform background removing in fingerprint image segmentation. Gaussian-Hermite moments can better separate image features based on different modes. We use Gaussian-Hermite moments of different orders to separate background and foreground of fingerprint image. Experimental results show that the use of Gaussian-Hermite moments makes a significant improvement for the segmentation of fingerprint images.

Sequential pattern mining is an important problem for data mining with broad applications. This paper presents a first-Horizontal-last-Vertical scanning database Sequential pattern Mining algorithm (HVSM). HVSM considers a database as a vertical bitmap. The algorithm first extends itemsets horizontally, and digs out all one-large-sequence itemsets. It then extends the sequence vertically and generates candidate large sequence. The candidate large sequence is generated by taking brother-nodes as child-nodes. The algorithm counts the support by recording the first TID mark (1-TID). Experiments show that HVSM algorithm can find frequent sequences faster than SPAM algorithm in mining the large transaction databases.

For multiple sequence alignment problem in molecular biological sequence analysis, a hybrid genetic algorithm and an associated software package called HGA-COFFEE are presented. The COFFEE function is used to measure individual fitness, and five novel genetic operators are designed, a selection operator, two crossover operators and two mutation operators. One of the mutation operators is designed based on the COFFEE’s consistency information that can improve the global search ability, and another is realized by dynamic programming method that can improve individuals locally. Experimental results of the 144 benchmarks from the BAliBASE show that the proposed algorithm is feasible, and for datasets in twilight zone and comprising N/C terminal extensions, HGA-COFFEE generates better alignment as compared to other methods studied in this paper.

Independent Component Analysis (ICA) is a useful statistical method for separating mixed data sources into statistically independent patterns. In this paper, we apply ICA to transform multivariate time series data into independent components (ICs), and then propose a clustering algorithm called ICACLUS to group underlying data series according to the ICs found. This clustering algorithm can be used to identify stocks with similar stock price movement. The experiments show that this method is effective and efficient, which also outperforms other comparable clustering methods, such as K-means.

Short sequences of system calls have been proven to be a good signature description for anomalous intrusion detection. The signature provides clear separation between different kinds of programs. This paper extends these works by applying fuzzy neural network (FNN) to solve the sharp boundary problem and decide whether a sequence is “normal” or “abnormal”. By using threat level of system calls to label the sequences the proposed FNN improves the accuracy of anomaly detection.

Some challenges for website designers are to provide correct and useful information to individual user with different backgrounds and interests, as well as to increase user satisfaction. Most existing Web search tools work only with individual users and do not help a user benefit from previous search experience of others. In this paper, a collaborative Web Mining System, Collector Engine System is presented, a multi-agent system designed to provide post-retrieval analysis and enable across-user collaboration in web search and mining. This system allows the user to annotate search sessions and share them with other users. The prototype system and component of Collector Engine System is discussed and described, and especially designs the web Agent, the knowledge discovery of web Agent is extracted based on a combination of web usage mining and machine learning. The system model is established and realized by J2EE technology. The system’s application shows that subjects’ search performances are improved, compared to individual search scenarios, in which users have no access to previous searches, when they have access to a limited of earlier search session done by other users.

Web page classification is one of the essential techniques for Web mining. This paper presents a framework for Web page classification. It is hybrid architecture of neural network PCA (principle components analysis) and SOFM (self-organizing map). In order to perform the classification, a web page is firstly represented by a vector of features with different weights according to the term frequency and the importance of each sentence in the page. As the number of the features is big, PCA is used to select the relevant features. Finally the output of PCA is sent to SOFM for classification. To compare with the proposed framework, two conventional classifiers are used in our experiments: k-NN and Naïve Bayes. Our new method makes a significant improvement in classifications on both data sets compared with the two conventional methods.